Research On Self-Anchored Cable-Stayed And Suspension Composite Bridge’s Finite Element Model Based On Sensitivity Analysis

The theoretical value of design parameters are always used to establish the initial finite element model in the structural design stage, but actually, the structural responses calculated by the initial finite element model often have certain differences compared with the test results. Finite element model updating is to analyze the static and dynamic responses of the structure in a variety of environments, and use the obtained data to correct the theory finite element model through the corresponding theory and optimization algorithm, so as to get a structure finite element model which can reflect the actual the structural responses more truly. The updated finite element model can not only reflect the actual state of the structure, but also can use it for the structural seismic calculation, wind resistant calculation and health monitoring, which can provide valuable, convincing and reliable data for the structural safety and reliability assessment. The thesis applies neural network to modify the finite element model of the trial model of Dongshaoxi bridge, the main work is as follows.Firstly, summarize the finite element model updating, introduce the model updating research situation at home and abroad, and put forward the research background and significance of this thesis.Secondly, the trial model of Dongshaoxi bridge and its construction process are introduced, and the finite element model of construction process is used for the partial design parameters sensitivity analysis. The bridge is a cable-stayed suspension composite system bridge.Compared with the conventional suspension bridge, the difference on sensitivity analysis is that the cable-stayed suspension composite system bridge considers the effect of cable initial tension in the construction process.Then, the characteristics, topological structure and learning rules of artificial neural networks are presented. BP and RBF neural network are two kinds of commonly used neural networks, and introduced mainly.Finally, the application of neural network used in model updating is introduced, the selection of sample uses uniform design method, and Sample data need to be preprocessed to ensure network efficiency and precision. The neural network toolbox of MATLAB is used for establishing the BP and RBF neural network of the bridge’s finite element model updating, and compared with the errors after training the two kinds of neural networks, then reaches that the RBF neural network to a higher precision.Therefore, the bridge’s finite element model updating is used the measured data and RBF neural network, and the results of updated model are in good agreement with the measured results.